Dangerous crash bug: SVGATextmode + X11 + Matrox Millenium II. Has anyone seen this?
(I apologize for the length of this message)
Versions:
svgatextmode: 1.8-5.2
xserver-svga: 3.3.2.3a-1
svgalibg1 1.3.0.19980916
Matrox Millennium II 8mb PCI video card
Awhile back, I loaded svgatextmode to try the 112x40 (very nice) text
mode that it provides. After that, I began to see a system-stopping
crash on occasion when running X. What is worse is the nature of the
crash itself. I have a newer 19" monitor (1600x1200 capable) that
fortunately has builtin protection circuitry against invalid timing
signals. After having run svgatextmode during bootup, I then run
'startx &' after logging in. About 35% to 45% of the time, running X
ends up with a window displayed on my monitor which says something like
the following:
Error: signal out of range
+125khz, +175hz
The window display comes from the monitor's own protection circuitry, it
is not from Linux. In addition, the keyboard is locked up (even numlock
doesn't toggle the numlock led), and Linux itself has crashed hard. I
found out later that the same problem happens with SVGALib programs. At
this point I have to use the hardware reset button to reboot the
computer. The bootup is a lengthy process because e2fsck wants to check
all the filesystems because they were not rebooted 'cleanly'. The worst
of this though, is the 'signal out of range' error, because it implies I
could have damaged my monitor, had I been using an old one without
protection circuitry.
Now to complicate matters further, I have learned from the
README.Debian file in svgatextmode a workaround for this problem. The
workaround in README.Debian refers to running the command 'cat /proc/pci
&> /dev/null' before using svgatextmode to solve a problem with PCI
based Matrox Millennium cards. In my case though, I can run
svgatextmode without difficulty and without the need of the above
workaround. However, if I bracket the invocation line for svgatextmode,
svgalib programs, and x11 with the above line, e.g.,
cat /proc/pci &> /dev/null
startx
cat /proc/pci &> /dev/null
I experience no more crashes. So, is it a svgatextmode problem
(svgatextmode was derived from svgalib IIRC), or is it an X11 problem,
or, could it be the kernel's PCI code? I'm not sure who I should be
talking to about this problem. I have not seen this kind of problem
mentioned in deb-user mailing list, or in comp.os.linux.x, or the bug
listings for the above programs on debian.org, so for all I know I'm the
only one seeing this behavior.
So, is there anyone out there with a Matrox M2 PCI card and a recent
monitor (with builtin protection) who is willing to try the above
software combination and see if the crash occurs on more than my
system? Please *don't do this* if you have an older monitor, I don't
want to be sued for having damaged it. :-) Install svgatextmode and
set it up for the 112x40 mode in TextConfig. Since you are using the
same card is I am, you'll probably need the slightly modified 80x25 mode
line below as the one in TextConfig doesn't work for us.
"80x25" 28.663 640 680 776 808 400 412 414 449 -Hsync +Vsync
font 9x16 # 31.532kHz/70.23Hz
With svgatextmode installed, start running X repeatedly. If its going
to crash, it will do so within the first ten or so executions.
Not knowing what to do with this problem, I've put it up on the Detain
User List, because of the seriousness of the bug; I feel compelled to
bring this up with somebody.
If there is a better place for this post, please let me know.
--
Ed C.
Reply to: